AI Voice For Demos, Thoughts?

[cosmicdolphin]

AI can mimic that, but it can't 'feel' the emotion or deliver the words from the heart.

It can because it's basically using performance capture in the same way real actors do for the performance of a CGI character in movies and then it's using that performace data to drive the model.

[Casey H]

There are many nuances of feeling, texture and color in a human singing voice. AI can mimic that, but it can't 'feel' the emotion or deliver the words from the heart.

True, but there already are quite a lot of synth voices, overly tuned voices for effect, etc. used in popular music today. In some genres, the difference in emotion might just be "cool" effect, not that different than we're hearing now.

[CTWF]

It could be that there is a physical limit for AI: By the time the AI is so advanced that it can produce music like a human, it will have to be nurtured for 5 (or 18) years and trained by so many humans or other AIs that it just is not cost efficient anymore. Not to mention the computing power needed. Possibly easier to raise a child and let it become a producer if it feels so.

[AlanHall]

It could be that there is a physical limit for AI: By the time the AI is so advanced that it can produce music like a human, it will have to be nurtured for 5 (or 18) years and trained by so many humans or other AIs that it just is not cost efficient anymore. Not to mention the computing power needed. Possibly easier to raise a child and let it become a producer if it feels so.

I, for one, would hate to spend 5 (or 18) years training an AI for music production only to have it decide it wants a career in marketing!

[Casey H]

It could be that there is a physical limit for AI: By the time the AI is so advanced that it can produce music like a human, it will have to be nurtured for 5 (or 18) years and trained by so many humans or other AIs that it just is not cost efficient anymore. Not to mention the computing power needed. Possibly easier to raise a child and let it become a producer if it feels so.

I, for one, would hate to spend 5 (or 18) years training an AI for music production only to have it decide it wants a career in marketing!

And then the bot will be promoted to management so it can tell the other bots they need to stay on schedule!

[CTWF]

I, for one, would hate to spend 5 (or 18) years training an AI for music production only to have it decide it wants a career in marketing!

Didn't we all disappoint our parents at some stage!

[mojobone]

I can absolutely get behind the idea of AI voices for demo purposes. How many times has Laskow played the demo from Rascal Flatts' "Here Comes Goodbye"? (and those guys didn't even beat the demo, IMO) There used to be a guy, sang all the Johnny Cash demos; I reckon he survives Johnny, but he's in need of gainful employment, since JR passed. With AI, he'd never have had the job in the first place, and might have gone into some field that could actually support a family and maybe raise another generation of hillbilly singers. But I digress. That demo for Rascal Flatts was deliberately written and arranged to fit the band, and it worked. An AI coulda maybe done that. (the demo, maybe not so much the vocal) Some songs have a more crooked path to being born. Spread your legs (or your cheeks, it's 2023 in the USA) and let the muse inseminate you. That'll beat a dang ol' calculator, every time.

[mojobone]

There are many nuances of feeling, texture and color in a human singing voice. AI can mimic that, but it can't 'feel' the emotion or deliver the words from the heart.

True, but there already are quite a lot of synth voices, overly tuned voices for effect, etc. used in popular music today. In some genres, the difference in emotion might just be "cool" effect, not that different than we're hearing now.

I think this is what's going to kill digital tuning.

[mojobone]

Is releasing a song in another artists voice a no-no? There are currently no laws that I am aware of that provide copyright or legal ownership of a voice. I'm sure it's just a matter of time though.

I think AI singing voices need to get better, but that too is only a matter of time. This stuff is progressing exponentially faster than previous advances in technology. Fun for sure!

Some commercial advertising producers have been successfully sued for violation of trademark, (not copyright) of name/image/likeness. You already can't put an AI-derived image of a sports star on a Wheaties box, if a release wasn't signed and an image was not (we need a better word than) photographed. If you call up Tom Waits to voice a commercial and he refuses to endorse your product, but then you go hire a soundalike (AI or otherwise) you are stepping in legal quicksand and advertisers already know this.